It's a brilliant book written by Sidney Dekker, a "Professor of Human Factors and Flight Safety". The basic point is that the default way of understanding bad outcomes is what he calls "the Old View or the Bad Apple Theory". He instead argues for the New View, where "human error is a symptom of trouble deeper inside a system".
Normally with a book like this, I read the first couple of chapters, say, "Ok, I get the idea," and can ignore the rest. After all, I both agree with and understand the basic thesis. But so far every chapter has been surprisingly useful; I keep discovering that I have Old View notions hidden away. E.g., when I discover a systemic flaw, I'm inclined to blame "bad design". But he points out that's a fancy way of calling the problem human error, just a different human and a different error than normal.
Even the driest parts are helped by his frequent use of examples, often taken from real-world aviation accident reports. There are also fascinating bits like a system for high-resolution markup of dialog transcripts to indicate timing (down to 1/10th second), speech inflection, and emphasis. I'll never use it myself, but I will definitely use the mindset that it requires.
Given how much time software projects spend dealing with bugs, I believe we need a new way to think about them, and for me this book describes a big piece of that.